Abstract
dc:description.abstractLithium-ion batteries underpin modern electric vehicles and grid storage, yet their safe operation depends on estimating internal states that cannot be measured directly. Conventional Kalman-based estimators presuppose accurate models and white Gaussian noise, assumptions that fail in practice because of modelling uncertainty, coloured noise, and temperature-dependent dynamics. This thesis develops Smooth Variable Structure Filter (SVSF) methods for robust battery state estimation, fault diagnosis, and thermal monitoring. A review of SVSF variants establishes their bounded-error robustness to modelling uncertainty. A third-order SVSF is then derived and its stability proven by Lyapunov analysis. Applied to a four-state lumped thermal model of a prismatic cell, it improves accuracy and suppresses chattering relative to existing variants and the Extended Kalman Filter (EKF). Because accuracy depends on model fidelity, rest time is examined as an overlooked factor in characterisation. Static and dynamic electrochemical impedance spectroscopy with distribution-of-relaxation-times analysis shows that rest periods alter impedance, explaining why standard pulse protocols yield models poorly suited to continuous operation. Pulsed charging is shown experimentally to generate more heat than continuous charging at equivalent throughput. Coupling SVSF estimation with cumulative-sum residual evaluation yields a diagnostic framework for current-sensor, voltage-sensor, and internal-resistance faults. It reduces detection time for current-sensor faults by an order of magnitude relative to EKF methods and isolates fault types through distinct residual signatures. Finally, state-of-charge estimation is addressed between −20◦C and −30◦C, where impedance rises sharply and model parameters become C-rate dependent. Modified test protocols and C-rate-dependent equivalent circuit models are developed. An Interacting Multiple Model framework incorporating SVSF attains mean errors below 1.4% and stays within 3% under ±30% parameter uncertainty, where the EKF diverges. An adaptive formulation embedding a series-resistance correction term in the state vector achieves mean errors below 1.1% at computational cost equivalent to the EKF.
Degree
thesis:*- Department dc:contributor.department
- Mechanical Engineering
- Year dc:date.issued
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Ebrahimizonouz, Farzaneh
- Advisors dc:contributor.advisor
-
- Habibi, Saeid
- Ahmed, Ryan
Rights
dc:rights- Statement dc:rights
-
- Attribution-NonCommercial-ShareAlike 2.5 Canada
- Licence dc:rights.uri
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Identifier URI
- https://doi.org/10.71548/25591